How shyamsridhar123/graphai works
Project type and purpose
• This is an open‐source, educational “intelligent knowledge graph” platform built in Python that uses Azure Cosmos DB (Gremlin API) for graph storage and Azure OpenAI for NLP/embedding services.
• Its purpose is to ingest natural‐language business events, extract entities/relationships over time (episodic ingestion), store them in a graph, then provide semantic search, analytics, and auto‐generated intelligence reports via CLI/demos (and future REST API).
File structure
• Top‐level directories:
• src/ – core GraphitiCosmos implementation (GraphitiCosmos class + asynchronous methods).
• demos/ – interactive and scenario demos (CLI‐style interfaces for exploratory use).
• scripts/ – data import, test‐data generation, environment setup helpers.
• data/ – sample and combined Manybirds e-commerce product/event datasets.
• docs/ – design docs, use‐case writeups, data‐loading guidance.
• visualizations/ – markdown with graph visualizations.
• intelligence_reports/ & ecommerce_intelligence_reports/ – generated BI outputs.
• Project config: pyproject.toml, uv.lock, .env, .gitignore.
• There's a single test file (test_search_enhanced.py) at root verifying search semantics.
README for additional insights
• The README details a layered, event‐driven architecture: Presentation, Business Logic, and Data layers.
• It lists key technologies: Azure Cosmos DB Gremlin API, Azure OpenAI (GPT-4 & embeddings), Python 3.12+, Gremlin Python driver, FastAPI (future REST API).
Diagramming conventions
• Related files under each component in small callouts (e.g., list demos/*.py under Presentation Layer).